English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Graphics Processing Units (GPUs) are widely-used accelerators fordata-parallel applications. In many GPU applications, GPU memory bandwidthbottlenecks performance, causing underutilization of GPU cores. Hence,disabling many cores does not affect the performance of memory-bound workloads.While simply power-gating unused GPU cores would save energy, prior worksattempt to better utilize GPU cores for other applications (ideallycompute-bound), which increases the GPU's total throughput.  In this paper, we introduce Morpheus, a new hardware/software co-designedtechnique to boost the performance of memory-bound applications. The key ideaof Morpheus is to exploit unused core resources to extend the GPU last levelcache (LLC) capacity. In Morpheus, each GPU core has two execution modes:compute mode and cache mode. Cores in compute mode operate conventionally andrun application threads. However, for the cores in cache mode, Morpheus invokesa software helper kernel that uses the cores' on-chip memories (i.e., registerfile, shared memory, and L1) in a way that extends the LLC capacity for arunning memory-bound workload. Morpheus adds a controller to the GPU hardwareto forward LLC requests to either the conventional LLC (managed by hardware) orthe extended LLC (managed by the helper kernel). Our experimental results showthat Morpheus improves the performance and energy efficiency of a baseline GPUarchitecture by an average of 39% and 58%, respectively, across severalmemory-bound workloads. Morpheus' performance is within 3% of a GPU design thathas a quadruple-sized conventional LLC. Morpheus can thus contribute toreducing the hardware dedicated to a conventional LLC by exploiting idle cores'on-chip memory resources as additional cache capacity.

Morpheus: Extending the Last Level Cache Capacity in GPU Systems Using Idle GPU Core Resources